[En-Nut-Discussion] SJA1000

[Henrik Maier]

I can follow Ole's point, there is an advantage in being able to use the
same Nut/OS build tree among different projects. 

But smaller code counts more, in particular on an AVR.

I also don't see the point of having the address run-time configurable while
the IRQ is fixed. Both parameters need to be set correctly and should be set
using the same method. To set the IRQ using nutconf and the address during
run-time is kind of inconsistent and possibly confusing.
 
I personally like the nutconf approach. You can configure and fine-tune
Nut/OS for a particular hardware or project without adding run-time code
overhead. Exactly what you want for a small embedded system.

Henrik


> -----Original Message-----
> From: Harald Kipp [mailto:harald.kipp at egnite.de]
> Sent: Friday, 31 August 2007 12:28 AM
> To: Ethernut User Chat (English)
> Subject: Re: [En-Nut-Discussion] SJA1000
> 
> Ole Reinhardt schrieb:
> > In this context I would vote for a new device registration framework
> > where you can pass the base address as well as the used interrupt and
> > other parameters to make the device code working better for different
> > platforms.
> >
> The disadvantage is, that the compiler will create larger and slower
> code. Specifically the powerful AVR bit set and bit clear instructions
> can be used with constant port addresses only. Thus, I'd suggest to
> support both methods:
> 1. Compile time configurable addresses for creating optimum code.
> 2. Run time configurable addresses for creating more flexible code.
> 
> Harald
> 
> _______________________________________________
> http://lists.egnite.de/mailman/listinfo/en-nut-discussion
>